Definition and validation of recompensation in patients with primary biliary cholangitis-related decompensated cirrhosis treated with ursodeoxycholic acid: Based on the BAVENO VII criteria

Study population

We analyzed 170 consecutive patients with PBC-related decompensated cirrhosis who were diagnosed and treated at the Xijing Hospital of Digestive Diseases (Xi’an, Shaanxi, China) from June 2006 to October 2023. PBC was diagnosed according to the European Association for the Study of the Liver.^[12]

The inclusion criteria were as follows: (1) standardized UDCA treatment; (2) meeting the imaging, radiological, or histological diagnostic criteria for cirrhosis; and (3) presence of ascites, hepatic encephalopathy, and/or variceal bleeding as the first presentation of decompensation.

The exclusion criteria were: (1) comorbid viral hepatitis, alcohol-related liver disease, drug-induced liver injury, autoimmune hepatitis, metabolic dysfunction-associated fatty liver disease, or hereditary liver disease; (2) follow-up time of less than 1 year; (3) treatment with glucocorticoids; (4) comorbid malignant tumors (excluding tumors that had been cured) or prior liver transplantation; and (5) dysfunction of vital organs, such as the heart, kidneys, and brain.

Study design

This was a retrospective cohort study. First, we identified the risk factors associated with liver transplant (LT)— free survival and defined recompensation criteria for PBC-related decompensated cirrhosis. Second, we evaluated the recompensation status in this group of patients treated with UDCA and analyzed the relation between recompensation and prognosis. Third, we explored dynamic changes in risk stratification and serum biochemical indicators between the recompensation and non-recompensation groups. Finally, we explored the risk factors associated with recompensation. The GLOBE risk score was used to estimate risk stratification and was calculated at baseline and in the first, second, and third years of follow-up.^[13] The primary outcome was liver-related death or LT. The secondary outcome was further decompensation, defined according to the BAVENO VII criteria,^[11] and jaundice was defined as total bilirubin (TBIL) > 3 mg/dL.^[14] Biochemical response was assessed using Paris I,^[6] Paris II,^[15] Rotterdam, ^[16] and Toronto^[17] at 1 year after enrollment. Separate retrospective analyses were performed using different response criteria. The study design was approved by the Ethics Committee of Xijing Hospital of the Air Force Military Medical University.

Treatment and follow-up

The baseline date was defined as the date on which the first decompensated event occurred during follow-up at our center. All patients with decompensated PBC received UDCA monotherapy at a standardized dose of 13 to 15 mg/kg/d. No escalation to second-line therapies was pursued owing to the high risk of adverse events associated with decompensated cirrhosis. Instead, these patients underwent intensified monitoring for disease progression (including the model for end-stage liver disease [MELD] score tracking and hepatic function tests every 3 months). Cirrhosis treatment was based on the standard of care according to the available recommendations at the time of accrual/follow-up. Concurrent medications were limited to the management of complications (including ascites and hepatic encephalopathy), with no immunomodulators or other hepatoprotective agents used.

Follow-up visits were based on clinical and laboratory assessments every 6-12 months or earlier (depending on clinical status), ultrasonography (screening for ascites) every 6-12 months, and upper endoscopy every 2-3 years in variance-free patients. Liver biopsies were analyzed according to the METAVIR scoring system^[18] and evaluated by two qualified and experienced pathologists who were blinded to the serological test results. Patients were classified as lost to follow-up if they missed three consecutive scheduled visits without contact. Patients lost to follow-up were excluded.

Statistical analysis

SPSS (version 26.0; IBM Corp., Armonk, NY, USA) and R statistical software 4.13 (Tsinghua) were used. The Shapiro–Wilk test was used to determine whether the data were normally distributed. Normally distributed continuous variables were expressed as mean ± standard deviation, and non-normally distributed variables were expressed as medians (interquartile range, IQR). Comparisons of the means of two continuous normally distributed variables were performed using the independent samples Student’s t-test, and the medians of two continuous non-normally distributed variables were compared using the Mann– Whitney U test. Cox regression analyses were used to examine the relation between prognostic indicators and LT-free survival, as well as the risk factors for recompensation. All variables with statistical significance in the univariate analysis were adjusted in the multivariate model. The DeLong test and C-index were used to compare the areas under the receiver operating characteristic curves. Statistical significance was set at P < 0.05.

Patient enrollment and follow-up characteristics

Figure 1 presents a flowchart of this study. A total of 291 patients with PBC-related decompensated cirrhosis were screened. After excluding 121 patients for various reasons, 170 were enrolled for a median follow-up of 39 months (IQR, 24-68 months), of whom 151 (81%) had ascites, one (1%) had hepatic encephalopathy, 18 (11%) had variceal bleeding, and 14 (8%) had both ascites and variceal bleeding at baseline. Patients underwent UDCA for a median of 26 months (IQR, 6-36 months) before enrolment.

Figure 1

Case screening process and study design. Recompensation was defined as the absence of decompensating events for 12 months following the last decompensated state and MELD <10. Standardized follow-up indicates that patients had 1-, 2-, and 3-year follow-up data available simultaneously. UDCA: ursodeoxycholic acid; PBC: primary biliary cholangitis; MELD: model for end-stage liver disease.

The baseline, 1-year, and outcome characteristics of the patients are presented in Table 1. Their mean age was 57 ± 9 years, with 89% (151/170) being women and 85% (144/170) being anti-mitochondrial antibody-positive. The median platelet (PLT), alkaline phosphatase (ALP), aspartate aminotransferase (AST), TBIL, and ALB values were 0.72 × lower limit of normal (LLN), 1.52 × upper limit of normal (ULN), 1.43 × ULN, 1.09 × ULN, and 0.91 × LLN. The median MELD and GLOBE scores were 12 (IQR, 10-14) and 1.40 (IQR, 0.84-2.09), respectively. The included patients exhibited low biochemical response rates, with only 48% (82/170) achieving a response according to the Paris-1 criteria. During follow-up, 28% (16/170) of the patients experienced primary outcomes, and 52% (89/170) experienced secondary outcomes.

Prognostic analysis at baseline and 1-year follow-up

At baseline, univariate analysis revealed that cholestasis indicators, such as ALP and gamma-glutamyl transpeptidase, were not associated with LT-free survival. Multivariate analysis demonstrated that age (hazard ratio [HR], 1.063; 95% confidence interval [CI], 1.017-1.112; P = 0.007), ALB level (HR, 0.009; 95% CI, 0.001-0.175; P = 0.002), and TBIL level (HR, 1.503; 95% CI, 1.191-1.898; P = 0.001) were associated with LT-free survival (Supplementary Table S1).

At the 1-year follow-up, univariate analysis showed that cholestasis indicators, such as ALP and gamma-glutamyl transpeptidase, and the Toronto criteria (ALP > 1.67 × ULN) were not associated with LT-free survival, whereas the Paris-1, Paris-2, and Rotterdam criteria were found to predict prognosis. Multivariate analysis demonstrated that AST (HR, 1.796; 95% CI, 1.114-2.896; P = 0.016), ALB (HR, 0.005; 95% CI, 0.001-0.114; P = 0.001), TBIL (HR, 1.343; 95% CI, 1.151-1.568; P < 0.001), and serum creatinine (HR, 4.436; 95% CI, 2.082-9.451; P < 0.001) were associated with LT-free survival (Supplementary Table S2).

ALP and TBIL levels are important indicators of PBC prognosis. To further explore the prognostic value of ALP, we stratified the included population according to TBIL values of 1.0 × ULN, 1.5 mg/dL, and 2.0 mg/dL at baseline and at the 1-year follow-up. As shown in Supplementary Table S3, the HRs of univariate Cox regression analyses were all less than one when TBIL was greater than 1.0 × ULN, 1.5 mg/dL, or 2.0 mg/dL, suggesting that elevated ALP may be a risk factor. Conversely, the HRs were all less than one when TBIL was lower than 1.0 × ULN, 1.5 mg/dL, or 2.0 mg/dL, suggesting that elevated ALP may be a protective factor. This was especially evident when we stratified the enrolled population according to a TBIL value of 1.0 × ULN at the 1-year follow-up, as the P values of both groups were less than 0.1.

Exploring the suitable recompensation criteria for PBC

Based on these findings, item A was defined as a biochemical response using the Paris-1, Paris-2, and Rotterdam criteria. We defined item B of recompensation as the absence of decompensating events from the last decompensated state in the following 12 months. ^[19] We defined item C of recompensation as MELD < 10 or Child-Pugh A.^[20] Given that ALP is a crucial marker for assessing biochemical response and considering the complex prognostic values of ALP in the enrolled patients, we also defined recompensation as items B and C.

Table 2 shows the effectiveness of various recompensation criteria in evaluating the primary outcomes. Compared to item A + item B + MELD < 10, the 3-year and 5-year areas under the receiver operating characteristic curves, as well as the C-index of item B + MELD < 10, were superior. Similarly, item B + Child-Pugh A showed the best results. The recompensation criteria, when combined with biochemical responses, did not enhance the predictive power of the primary outcomes defined by item B + MELD < 10 or item B + Child-Pugh A.

Compared to item B + Child-Pugh A, item B + MELD < 10 improved the predictive power of the 3-year primary outcomes (Table 2), and the combination of item B + MELD < 10 and item B + Child-Pugh A did not improve the predictive power of the primary outcomes established by item B + MELD < 10.

Therefore, we defined recompensation as items B (absence of decompensating events from the last decompensated state in the following 12 months) and C (MELD score < 10) in patients with PBC-related decompensated cirrhosis.

Recompensating status and prognosis

Forty-five patients achieved recompensation at a median of 26 months (IQR, 23-37 months) after enrolment. The recompensation rate of the included patients was 26% (45/170). The primary and secondary outcome rates were 2% (1/45) and 22% (10/45) in the recompensation group and 22% (27/125) and 63% (79/125) in the nonrecompensation group, respectively. Significant differences were observed in primary and secondary outcome rates between the recompensation and non-recompensation groups (P < 0.05). Cox analyses revealed a substantial increase in the hazards of liver-related death or LT attributable to non-recompensation (HR, 16.48; 95% CI, 2.23-121.57; P = 0.006; Figure 2).

Figure 2

Kaplan-Meier plots of LT-free survival in patients with PBC-related decompensated cirrhosis, stratified by recompensation status. Recon, recompensation. LT: liver transplant; HR: hazard ratio.

Dynamic changes in risk stratification and serum biochemical indicators

Our study cohort included 71 (42%) patients with concurrent 1-, 2-, and 3-year follow-up data. A previous study confirmed that using GLOBE values of 0.5 and 1.8 could stratify patients with PBC into low-, medium-, and high-risk groups. 22 Accordingly, we plotted Sankey diagrams to represent changes in risk stratification using GLOBE scores from baseline to 36 months. These diagrams depicted a remarkably greater reduction in the proportion of high-risk patients in the recompensation group than in the non-recompensation group (Figure 3, Supplementary Table S4). Although both groups exhibited improvements in ALP and AST levels during the 36-month treatment period, the recompensation group demonstrated a significantly higher proportion of stable improvements in PLT, ALB, TBIL, and international normalized ratio (INR) levels (Figure 4).

Figure 3

Sankey diagrams illustrating changes in risk stratification from baseline to 36 months. Sankey diagrams represent major transitions in patient risk status. Colors indicate different GLOBE classifications: yellow for low risk, pink for medium risk, and red for high risk. Column lengths represent patient proportions in each risk category. Thicker lines indicate a greater number of patients transitioning between categories. Patients with PBC were stratified as low, medium, or high risk based on GLOBE values of 0.5 and 1.8. (A) Total patients (n = 71). (B) Patients with recompensation (n = 33). (C) Patients without recompensation (n = 38).

Figure 4

Dynamic changes in serum indicators in patients with and without recompensation over 36 months of treatment. Median values and IQRs are shown. Differences between groups at each time point were assessed using Mann-Whitney U tests. Recon, recompensation. PLT: platelet; AST: aspartate aminotransferase; ALB: albumin; TBIL: total bilirubin; ALP: alkaline phosphatase; INR: international normalized ratio; mo: months; ULN: upper limit of normal; LLN: lower limit of normal; IQR: interquartile range; NS: not significant. ^*P < 0.05, ^**P < 0.01, ^***P < 0.001.

Baseline indicators, one-year biochemical criteria, and other parameters related to recompensation

As shown in Supplementary Table S5, baseline indicators, including sex, age, PLT, ALB, TBIL, and prothrombin time (PT), were associated with recompensation. Multivariate analysis demonstrated that PLT (HR, 1.688; 95% CI, 1.184-2.349; P = 0.003) and TBIL (HR, 0.517; 95% CI, 0.306-0.873; P = 0.014) were independently associated with recompensation. At the 1-year follow-up, univariate analysis indicated that the Paris-1, Paris-2, and Rotterdam criteria were associated with recompensation. Multivariate analysis demonstrated that the Rotterdam criteria (HR, 2.433; 95% CI, 1.350-4.421; P = 0.003) were independently associated with recompensation.

A total of 106 (62%) patients underwent liver biopsy during the follow-up period, including 15 (14%) with severe interface hepatitis and seven (7%) with mild to moderate interface hepatitis. Chi-square test analysis revealed that neither mild to moderate interface hepatitis nor severe interface hepatitis was associated with recompensation (P = 0.067 and 0.313, respectively). However, Cox analysis revealed that severe interface hepatitis was associated with a decreased likelihood of recompensation (HR, 3.055; 95% CI, 1.261-7.399; P = 0.013).